Vision Autonomes Fahren im Urbanen Umfeld - Dr. Heinrich Gotzig g Expertise & RAISE Director Valeo CDV Valeo Senior Expert - REM 2030
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Vision Autonomes Fahren im Urbanen Umfeld Dr. Heinrich Gotzig g Expertise & RAISE Director Valeo CDV Valeo Senior Expert
Agenda Automotive Megatrends On the way towards Autonomous Driving Sensor Technologies Fusion concept / architecture Park4U®Remote Current development activities Summary 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Pleasure to drive We enjoy driving… Here. There? If d driving g is s a bu burden, de , what at is s the t e end e d user use expectation? e pectat o 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Urbanization More and more driving takes place in urban environments 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Megatrends: Sustainable Mobility Mobility in urban environments Î Parking and maneuvering support Î Automatic docking and charging for EVs Î Car-sharing with partially autonomous elements, e.g. redispatching Stay mobil at an older age Î Not enoug adults to take care on the eldery Î Quality of living Accident Avoidance: Î „Vision Vision Zero“ Zero stop traffic fatalities – Active safety systems – Intelligent infrastructure Î „ „Decade for Action“: Reduction of worldwide traffic fatalities by 5 million 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
On the way to Autonomous Driving Routine Ausnahme unaufmerksam f k positives üb f d t überfordert Fahrerlebnis Delegation/ unterstützender Warnung angepasste Eingriff Already existing: Information Semi-automatic Parking High-beam Assist Lane Keeping Emergency Braking Active projects: The Vision: Autopilot AMARE ICADAC SPEEDCAM AROS Modulowatt odu o a Vehicle dynamics Automatic docking/parking Improved vision in Next generation for automation in for Electric Vehicles adverse weather Traffic Sign Recognition Parking & Manoeuvring Fraunhofer 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
On the way to autonomous Driving Concrete timeline communicated by automotive OEMs: “Latest in 2020, technologies like automatic driving will be standard.” Ulrich Hackenberg VW, Member of the Board for R&D “Vehicles that partially drive themselves will be available by the middle of the p decade with more sophisticated self- driving systems by the end of the decade” Alan Taub GM VP Global Research & Development 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Autonomous Driving – Technical Challenges Environmental Sensors Signal Processing Today: Eye of Driver Today: Human brain Aut Dri: several ranging + vision sensors, Aut Dri.: signal processing unit which which are able to record ALL needed p computes all recordes informations informations on environment (sensors + infrastructure) and decides on activities like perfect driver would do Infrastructure User Today: passive traffic signs Today: driver in full controll of car Aut Dri: active traffic signs (time dependent) Aut Dri.: driver has to hand over / get back informations on traffic situations / controll to / of vehicle. Acceptance, under participants, GPS, maps, … which conditions, situations, how … ??? Diff Different t user expectations t ti (age, ( culture, lt …)) 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Where is Valeo today? Valeo offers Multiple Sensor Technologies Ranging and perception necessary to support both models Î Vision systems as reference to what the human perceives – Also provide „every day value“ through enhanced visibility Î Radar, Lidar, IR and Ultrasonics to complement the human senses Sensor fusion takes care of functional safety through redundancy Î Odometry experience from multiple OEM projects in the last 5 years Î Fusion map in series development Technology Strengths Limitations Market Trend Ultrasonics Inexpensive (150m) Significant cost (100-200€) FMCW Good angular accuracy (up to 1°) Limited object separation and Ò Compact size (e.g. 7x7x5cm) classification Radar 79GHz – Much better resolution than FMCW Regulations not yet approved UWB Ò Camera – monocular Multifunctional, close to human perception Same as human vision (rain/fog etc) (i (image processing) i ) R Reasonable bl costt (60-100€) (60 100€) N d illumination Needs ill i ti att night i ht Ò Camera – stereo Multifunctional, close to human perception Same as human vision (rain/fog etc) Calculates depth/distance information Needs illumination at night ? Camera – 3D TOF Multifunctional, close to human perception Same as human vision (rain/fog etc) (PMD) Calculates depth/distance information Resolution limited by prototype chips ? Lidar – Fixed Beam Precise „in/out“ localization of objects Similar to human vision (rain/spray) Reasonable cost (40-60€) Sensor needs to be visible Ò Lidar – Scanning Long-range capable (>100m) Similar to human vision (rain/spray) Very good angular accuracy (0,25 (0 25°)) Sensor needs to be visible Ò 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Valeo Architecture Proposal Fusion ECU as Core Element C2 R1 U7 U1 U8 U2 U9 U3 LS C4 FC Fusion C1 R3 ECU U10 U4 U11 U5 U12 U6 C3 R2 Front Camera Viewing Camera Ultrasonic Sensor Radar Sensor Laser Scanner Fusion Controller 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Map-based Detection of the Environment Basis for Sensor Data Fusion Combination of sensor data from different sources Î Redundancy for safety classification Î Increased density of information in difficult conditions Currently still separate HMI Four Fusion Layers Valeo Fusion-ECU Î Points, Features, Objects and Scenarios UPA Sensors Î Common path planning, tracking and prediction P4U Sensors Open Architecture Wide-angle Digital Camera Î Integration of 3rd party sensor information Î Scalable for different models Distributed Architecture Option Î Stand alone camera and Park4U can share the map Stand-alone Î Identical software modules for different architectures 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Fusion concept Associates features from different sensor channels Associates objects from different sensor channels Computes a combined confidence of a feature Computes a combined confidence of a object Fuses the location information and computes a new uncertaintyy 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Modular Feature/Technology Approach Standard (Technology Bricks) Customer Specific Implementation Package/ Mechanical Brackets Integration Ultrasonic Interface Path Detection Sensors HW/SW Planning Physical Electrical Layer Integration Tracking Odometry Software NOS Integration Vision Interface Classi- … Sensors HW/SW fication System HMI I t Integration ti Threat … Assessmt. Interface Steering HW/SW System Radar Interface Decision … Sensors HW/SW Module Interface Braking HW/SW System Sensor Frontend Core Algorithms Vehicle/System Integration 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
360Map Fusion Software Architecture High Level Overview S S SYSTEM OUTPUT O Dot Matrix Video Steering Brake Wheellspeed Speakers Display Screen Actuator Actuator Gearbox ector Sensors Vehicle Dynamics Sele HMI Frontend (Audio, Video, Haptics) Frontend Feature Logic and Abstraction ng CLE STAUS S cs Modellin Steering Steering Warning and Vehicle Torque Path Angle DRIVER IINPUTS Information Control Driver Intention Planning Interpretation layer OST VEHIC e Dynamic Object layer Odometry MAP (History, Prediction, Acceleration Contour layer Tracking) aw Rate witches Points layer HO Vehicle Sw Environmental Information Ya (Point / Contour / Object) Ultrasonic Sensor control and low Image Core Module level processing Processing … .. Ultrasonic Infrared Radar Camera Raw Signal Sensor Frontend Frontend Frontend Frontend (other) Processing Diagnosis Sensor Control Loop ENVIRONMENT DETECTION 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Parking/Manoeuvring and Urban Driving ongoing development activities Ultrasonic Camera Radar Lidar Parallel Perpendicular Turn Perpendicular Passage Lateral Blind Spot A Parking B Parking C (Side Protection) D Parking E F (Reverse) (Forward) in/ in/ in/ out out out Visibility – Driver Information Rear View Rear/U-View Top/U-/O-View Top/U-/O-View Top/U-/O-View Top/U-/O-View Available (2004) Available (2004) Available (2009) Available (2010) Available (2010) Available (2010) Escalation – Driver Warning Ultras. Blind Spot/ Video Overlays Video Overlays FlankGuard Video Overlays PathVue Door Opening Alert [UPark – on(2007) Available hold] Available (2007) Planned (2010) Available (2010) Planned (2013) Available (2011+) Park4U Park4U Active Intervention – Steering Side Collision Prev. Gen2: Stop@End Gen3: Line Detection FlankGuide Garage Parking PathGuide Parallel Unpark Available (2010) Available (2010+) Available (2013+) Planned (2013+) Planned (2015) Planned (2013+) 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Ultrasonic/Camera Fusion Example Park4U® Remote 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Vienna Convention on Road Traffic – Park4U ®Remote Driver has to consciously trigger the maneuver Î Process is predefined within close limits (speed, permissible movements) Driver has to confirm the maneuver is proceeding properly Driver has the opportunity to survey the vehicle surroundings Î Better than from the inside… Any ambiguity will interrupt the maneuver immediately Î Sensors monitor surrounding and will brake on unexpected obstacles 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Park4U®Remote next steps 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
Summary One automotive megatrend is mobility in urban environment A t Autonomous d driving i i needs d to t be b reliable, li bl affordable, ff d bl user friendly f i dl ... Systems require redundant information Î Sensor data from ranging systems (Radar, Ultrasonic) are fused with Vision systems using image processing for object detection Î Viewing cameras become sensors Environment map as fusion platform Î Object detection by multiple sources (confidence, classification etc.) Î Calculation of vehicle position and trajectory Î Points layer Æ Contour layer Æ Object layer Æ Interpretation layer Valeo is working towards autonomous driving in urban environment 13.Nov. 2012 „Zulieferertag Automobil“ VALEO confidental I
You can also read